CN108052980A - Air quality grade detection method based on image - Google Patents
Air quality grade detection method based on image Download PDFInfo
- Publication number
- CN108052980A CN108052980A CN201711389016.6A CN201711389016A CN108052980A CN 108052980 A CN108052980 A CN 108052980A CN 201711389016 A CN201711389016 A CN 201711389016A CN 108052980 A CN108052980 A CN 108052980A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msubsup
- mtd
- image
- mtr
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/85—Investigating moving fluids or granular solids
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/85—Investigating moving fluids or granular solids
- G01N2021/8578—Gaseous flow
Abstract
The invention discloses a kind of air quality grade detection methods based on image, belong to technical field of image processing.This method does color space conversion to the data set of collection first, obtains the half-tone information and colour information of image.Then extract the gray channel of the multiple scales of image and color channel local message entropy, and by the use of local message entropy average and slope as feature.Classify followed by SVM classifier to image according to air quality grade, and then obtain air quality grade detection model.Finally complete air quality grade detection.Go out the air quality grade estimation effect of method given in the present invention better than the prior art, be of great significance to air quality monitoring in daily life.
Description
Technical field
It is particularly a kind of based on image the present invention relates to a kind of method by image detection image air quality grade
Air quality grade detection method.
Background technology
Air quality grade is to weigh an important indicator of air quality, is broadly divided into 6 grades:Excellent, good, slight dirt
Dye, intermediate pollution, serious pollution, serious pollution, higher grade are bigger to human health damage.At present, air quality grade is
It is determined according to 6 kinds of main pollutant consistences, it is then main to 6 kinds dirty first by accurate apparatus measures pollutant concentration
Air quality graduation is calculated in dye object respectively, finally selects maximum determine from the air quality graduation of various pollutants
For air quality grade, the maximum pollutant of air quality graduation is determined as primarily when excellent when air quality grade is more than
Pollutant.The detection of main pollutant consistence is required for relying on precision instrument, these instrument costs are higher and need periodic maintenance.
By it has been observed that the image taken in the case of different air quality grade, there are notable differences in quality.Therefore want to design
Go out a kind of method of image detection air quality grade.It, should there is presently no the method by image detection air quality grade
Method may be considered a kind of extension application of non-reference picture quality appraisement, so attempting non-reference picture quality appraisement phase
Pass technology is applied in this method.Classical non-reference picture quality appraisement method was broadly divided into two categories below in recent years:
(1) based on traditional machine learning method.Manually extraction feature trains to obtain no ginseng by traditional machine learning
Examine image quality evaluation model.The machine learning method that this kind of method generally uses has SVM, random forest etc., can distinguish figure
As the characteristics of image of quality difference has very much, such as:It has been fitted the statistical property of image using generalized Gaussian distribution, and will obtain
Parameter is as feature;The local message entropy of image is extracted as feature;It is extracted DCT coefficient Kurtosis values, anisotropy entropy
Maximum value tag;By the use of information such as image gradients as feature;It is extracted colored degree, sharpness and the contrast metric of image;
Using the non-Gaussian system of NSS features, local dependence and exponential damping characteristic;Part NSS features are extracted, are learnt by rarefaction representation
The quality perceptual filter that algorithm obtains carries out feature coding;
(2) it is based on deep learning method.This kind of method is divided into two types, and a kind of is that the feature extracted is inputted nerve
In network, such as:The visual signatures such as edge amplitude, edge length, background liveness and background luminance are input to individual layer feedforward god
Through in network;The features such as phase equalization image average, entropy and image gradient are input in generalized regression nerve networks.It is another
Kind is directly to input an image into neural network framework, and carrying out feature learning by frame obtains non-reference picture quality appraisement mould
Type.
At present, the air quality situation in entire city detects by setting up several monitoring stations in each city in the country, respectively
A monitoring station staff detects the concentration of six kinds of pollutants in air using various instruments daily and is converted to corresponding air
Credit rating is published on official website.The purchase and periodic maintenance of instrument so that air quality monitoring cost is higher, and by setting up
The air quality that the mode of monitoring station detects entire city is coarseness, it is impossible to the air of reflection city every nook and cranny well
Quality.
The content of the invention
It is an object of the invention to provide a kind of new methods based on image detection air quality grade.
The technical solution achieved the object of the present invention is:A kind of new side based on image detection air quality grade
Method comprises the following steps:
The original image set of collection is transformed into Lab space by step 1 from rgb space;
Step 2, the spatial domain local message entropy for extracting image L passages;
Step 3 does block DCT transform to the image for being converted into Lab space;
Image carries out the extraction of frequency domain local message entropy after step 4, the block DCT transform obtained to step 3;
Step 5 carries out mean value computation and slope fit processing to the local message entropy obtained in step 2 and step 4, obtains
Corresponding average and slope;
Step 6 carries out the original image set of collection 2 times of down-sampling operations, and the image set after down-sampling is empty from RGB
Between be transformed into Lab space, repeat operation of the step 2 to step 5 afterwards, obtain corresponding average and slope;
Step 7 carries out the original image set of collection 4 times of down-sampling operations, and the image set after down-sampling is empty from RGB
Between be transformed into Lab space, repeat operation of the step 2 to step 5 afterwards, obtain corresponding average and slope;
Step 8, by the average extracted in step 5, step 6 and step 7 and slope characteristics and its corresponding air quality etc.
Grade label is input to SVM classifier and models to obtain Detection of Air Quality model simultaneously;
Step 9, the model obtained with step 8 are detected air quality grade.
Compared with prior art, the present invention its remarkable advantage is:1) present invention easily can quickly detect air matter
Grade is measured, it is the air quality grade that can obtain periphery that people, which need to only utilize the equipment such as mobile phone to upload piece image,;2) it is of the invention
The shortcomings that compensating for the prior art, the air quality grade detection method that the present invention realizes have fine granularity, the characteristic of low cost;
3) the air quality grade detection model estimated accuracy that the present invention obtains is higher, can be very good the sky that people is helped to detect periphery
Makings amount;4) present invention can be good at that people's stroke of making rational planning for is helped to avoid injury of the air pollution to body;5) this hair
It is bright to introduce classical non-reference picture quality appraisement technology and be applied to after making improvements in air quality grade detection,
It realizes through image to detect air quality grade.
The present invention is described in further detail below in conjunction with the accompanying drawings.
Description of the drawings
Fig. 1 is the flow chart of the air quality grade detection method the present invention is based on image.
Fig. 2 is the training set that web crawlers crawls and the schematic diagram of corresponding label, wherein, figure (a) is that air quality is
The image taken when excellent;Figure (b) is the image taken when air quality is good;Figure (c) is that air quality is slight pollution
When the image that takes;Figure (d) is the image taken when air quality is intermediate pollution;Figure (e) is that air quality is severe
The image taken during pollution;Figure (f) is that air quality is the image taken during serious pollution;
Fig. 3 is piece image color space conversion schematic diagram, wherein, figure (a) is the image of rgb space;It is Lab to scheme (b)
The image L passages in space;Figure (c) is the image a passages of Lab space;Figure (d) is the image b passages of Lab space.
Fig. 4 is six kinds of different air quality grade image L channel spaces domain local message entropy set normalization histograms.
Fig. 5 is the schematic diagram after tri- passage block DCT transforms of piece image L, a, b, wherein figure (a) (b) is that image L leads to
The front and rear figure of road block DCT transform;Figure (c) (d) is the front and rear figure of image a passages block DCT transform;It is image b passages to scheme (e) (f)
The front and rear figure of block DCT transform.
Fig. 6 is histogram, and (a) (b) (c) represents six kinds of different air quality grade image L, a, b passage frequency domains respectively
Local message entropy normalization histogram.
Fig. 7 is six kinds of different air quality grade image local message entropy average and slope distributions under archeus, wherein,
It is L channel spaces domain local message entropy average and slope distribution to scheme (a);Figure (b) be L passage frequency domain local message entropy averages and tiltedly
Rate is distributed;It is a passage frequency domain local message entropy averages and slope distribution to scheme (c);It is that b passage frequency domain local message entropys are equal to scheme (d)
Value and slope distribution.
Further detailed description is done to the present invention below in conjunction with the accompanying drawings.
Specific embodiment
With reference to Fig. 1, one kind of the invention is based on image detection air quality grade method, comprises the following steps:
The original image set of collection is transformed into Lab space by step 1 from rgb space;Specially:
Step 1-1, image is transformed into XYZ space from rgb space, conversion formula is:
Wherein, R, G, B distinguish the R passages, G passages, channel B of representative image, and X, Y, Z represent the X passage of image, Y respectively
Passage, Z passages;
Step 1-2, image is transformed into Lab space from XYZ space, conversion formula is:
L=116f (Y/100) -16
A=500 [f (X/95.047)-f (Y/100)]
B=200 [f (Y/100)-f (Z/108.883)]
Wherein, t represents X, Y, Z passage, and f (t) is the transfer function from XYZ space to Lab space, and L, a, b are represented respectively
L passages, a passages, the b passages of image.
Step 2, the spatial domain local message entropy for extracting image L passages;Specially:
Step 2-1:Piecemeal processing is done to image L passages, local block size is 8*8;
Step 2-2:The comentropy of L path partially blocks is calculated, the calculation formula of comentropy is:
E=- ∑sxp(x)log2p(x)
Wherein, x refers to the value of pixel in L path partially blocks, and p (x) refers to that pixel value is x's in L path partially blocks
Probability, E refer to localized mass information entropy;
Step 2-3:All spatial domain local message entropys of L passages according to comentropy size are ranked up, are expressed as:
Wherein, n is spatial domain L path partially block numbers,The spatial domain letter of n localized mass of L passages of finger
Cease entropy andSslRepresent the spatial domain L path partially comentropy set after sequence.
Step 3 does block DCT transform to the image for being converted into Lab space;Specially:
Localized mass dct transform is done respectively to image L, a, b passage, the calculation formula of dct transform is:
Wherein, f (i, j) is the pixel value that spatial domain L, a, b path partially block internal coordinate is (i, j), and local block size is
N*N takes 8*8, and (u, v) is the coordinate of pixel after L, a, b path partially block dct transform, and F (u, v) refers to L, a, b passage office
The DCT coefficient obtained after portion's block dct transform.
Image carries out the extraction of frequency domain local message entropy after step 4, the block DCT transform obtained to step 3;Specially:
Step 4-1, piecemeal processing is done to image L, a, b passage after block DCT transform, local block size is 8*8;
Step 4-2, the comentropy of image L, a, b path partially block after block DCT transform is calculated, the calculating of comentropy is public
Formula is:
E=- ∑sxp(x)log2p(x)
Wherein, x refers to the value of pixel in image L, a, b path partially block after block DCT transform, and p (x) refers to single-pass
Pixel value is the probability of x in road localized mass, and E refers to localized mass comentropy;
Step 4-3:It can be represented respectively after sorting to all frequency domain local message entropys of L, a, b passage according to comentropy size
For:
Wherein, n is frequency domain L, a, b path partially block number;The frequency domain of n localized mass of L passages of finger
Comentropy andSflRepresent the frequency domain L path partially comentropy set after sequence;Refer to
Be n localized mass of a passages frequency domain information entropy andSfaRepresent the frequency domain a path partiallies after sequence
Comentropy set;Refer to the frequency domain information entropy of n localized mass of b passages andSfb
Represent the frequency domain b path partially comentropy set after sequence.
Step 5 carries out mean value computation and slope fit processing to the local message entropy obtained in step 2 and step 4, obtains
Corresponding average and slope;Specially:
Step 5-1, S is taken respectivelysl、Sfl、Sfa、Sfb60% local message entropy, value among four local message entropy set
Mode is:
Step 5-2, respectively to SSl_60%、SFl_60%、SFa_60%、SFb_60%Four local message entropy set carry out mean value computation
And slope fit, the average and slope calculation formula of local message entropy set are:
Wherein, S represents SSl_60%、SFl_60%、SFa_60%、SFb_60%, m is the size of local message entropy set S, meanSIt is office
The average of portion comentropy set S, skewSRepresent the slope of local message entropy set S;
Step 5-3, the corresponding average of four local message entropy set and slope are expressed as:
F1=[meanSl_60%, skewSl_60%, meanFl_60%, skewFl_60%, meanFa_60%, skewFa_60%,
meanFb_60%, skewFb_60%]。
Step 6 carries out the original image set of collection 2 times of down-sampling operations, and the image set after down-sampling is empty from RGB
Between be transformed into Lab space, repeat operation of the step 2 to step 5 afterwards, obtain corresponding average and slope is:
Step 7 carries out the original image set of collection 4 times of down-sampling operations, and the image set after down-sampling is empty from RGB
Between be transformed into Lab space, repeat operation of the step 2 to step 5 afterwards, obtain corresponding average and slope is:
Step 8, by the average extracted in step 5, step 6 and step 7 and slope characteristics and its corresponding air quality etc.
Grade label is input to SVM classifier and models to obtain Detection of Air Quality model simultaneously;The SVM classifier is by LibSVM works
Tool bag is realized.
Step 9, the model obtained with step 8 are detected air quality grade.
Air matter is applied to invention introduces classical non-reference picture quality appraisement technology and after making improvements
It measures in grade detection, realizes through image to detect air quality grade.
With reference to embodiment, the present invention will be further described in detail.
Embodiment
The system invention is detected using the natural image of outdoor acquisition as input using image processing means by input picture
The air quality grade of image taking there and then.In order to detect the performance of the system invention, first with web crawlers from rich
100 width images are crawled on visitor as data set, by 100 width images according to 8:2 ratio is randomly divided into training set and test set.So
Afterwards, air quality grade estimation model is built on training set with the present invention, finally utilizes the feature of present invention extraction test set
The air quality grade of test set image taking there and then, statistical estimate air quality can be estimated by being input in estimation model
The correlation of grade and real air credit rating calculates the air quality grade estimation model accuracy of the system invention structure.
The natural color image size that the flow of the present embodiment as shown in Figure 1, swashed by web crawlers from blog is got
For 640 × 416, image totally 100 width crawls image air quality grade label corresponding with its as shown in Fig. 2, can be with from Fig. 2
Find out as air pollution aggravates, for image by clearly fogging, color becomes grey by blueness.
Color space conversion is done to 100 width images, Lab space is transformed into from rgb space and extracts three scale skies of L passages
Between tri- passages of domain local message entropy and L, a, b three scale frequency domain local message entropys.It concretely comprises the following steps:
The first step:Lab space is transformed into from rgb space using 100 width image of matlab function pairs, is one as shown in Figure 3
Width image color space transition diagram, wherein, figure (a) is the coloured image of rgb space, and each scenery in image has it
Distinctive color, if sky is blue, building is white.
Second step:Utilize formula E=- ∑sxp(x)log2P (x) extraction L channel spaces domain local message entropy, local block size
For 8*8, obtained after L channel spaces domain local message entropy set is sorted according to the size of local message entropyIt is that the corresponding image L channel spaces domain of six kinds of difference air quality grades is locally believed as shown in Figure 4
Cease entropy set normalization histogram.
3rd step:Block DCT transform, local block size are carried out using tri- passages of L, a, b of matlab function pair images
It is the schematic diagram before and after tri- passage block DCT transforms of piece image L, a, b as shown in Figure 5 for 8*8.
4th step:E=- ∑s are utilized to tri- passages of L, a, b after block DCT transformxp(x)log2P (x) extraction frequency domains office
Portion's comentropy, local block size is 8*8, by tri- passage frequency domain local message entropy set of L, a, b according to local message entropy
It is obtained after size sequence:
It is corresponding image L, a, b passage the frequency domain local message of six kinds of difference air quality grades as shown in Fig. 6 (a)-(c)
Entropy normalization histogram is distributed.
5th step:100 width image, 2 times of down-samplings are repeated into step 1 to step 4 and obtain four local message entropy set.
6th step:100 width image, 4 times of down-samplings are repeated into step 1 to step 4 and obtain four local message entropy set.
Mean value computation is carried out to the four local message entropy set extracted under 100 width image, three scales and slope is intended
It closes, is six kinds of different air quality grade images corresponding local message entropy average and slope point under archeus as shown in Figure 7
Cloth.It concretely comprises the following steps:
The first step:Among the local message entropy set of image 60% local message entropy is taken under each scale, is denoted as:
Second step:To SSl_60%、SFl_60%、SFa_60%、SFb_60%Four local message entropy set carry out respectively mean value computation and
Slope fit, formula are respectively:
3rd step:After mean value computation and slope fit being carried out to four local message entropy set of three scales of each image
Three 8 dimensional features are obtained, are denoted as:
F1=[meanSl_60%, skewSl_60%, meanFl_60%, skewFl_60%, meanFa_60%, skewFa_60%,
meanFb_60%, skewFb_60%]
The feature of 100 width images has been extracted, has then been modeled to obtain air quality grade detection model with SVM classifier.This
100 width of instance graph image set estimates the air quality grade of test set image taking there and then, i.e., using 1000 cross validations
The image of random selection 80% is modeled as training set every time, and the image of residue 20% is tested as test set, carries out 1000 altogether
It is secondary.It concretely comprises the following steps:
The first step:Corresponding three 8 dimensional features of 80 width images and real air quality grade label are input to SVM points
In class device (interface for calling LibSVM kits), air quality grade estimation model is obtained.
Second step:The air quality grade that corresponding three 8 dimensional features of 20 width images and the first step are obtained estimates model
(interface for calling LibSVM kits) is input in SVM classifier, obtains 20 corresponding air quality grade estimates.
Table 1 is the experiment pair of the method for the present invention and existing methods non-reference picture quality appraisement method on notebook data collection
Than being particularly suited for air quality grade detection by comparing the discovery present invention.
Table 1
Algorithm | LCC | SROCC | TIME(s) |
SSEQ | 0.8453 | 0.8412 | 1345.290152 |
IQALE-a | 0.8600 | 0.8344 | 1892.137452 |
IQALE-b | 0.8620 | 0.8432 | 1912.642230 |
IQALE-a, b | 0.8540 | 0.8319 | 2827.984681 |
Our-a | 0.8641 | 0.8377 | 1743.231679 |
Our-b | 0.8654 | 0.8458 | 1776.038448 |
Our-a, b | 0.8807 | 0.8608 | 2397.254333 |
As known from Table 1:The air quality grade estimation model that the present invention obtains can accurately estimate that Image Acquisition is worked as
When local air quality grade, correlation LCC, SROCC between air quality grade estimate and actual value is better than it
His non-reference picture quality appraisement method, and this method decreases modeling consumption while highest estimated accuracy is obtained
When.The shortening of time and the daily life air quality grade monitoring that rises to of estimated accuracy provide a convenient.
Claims (9)
1. a kind of air quality grade detection method based on image, which is characterized in that comprise the following steps:
The original image set of collection is transformed into Lab space by step 1 from rgb space;
Step 2, the spatial domain local message entropy for extracting image L passages;
Step 3 does block DCT transform to the image for being converted into Lab space;
Image carries out the extraction of frequency domain local message entropy after step 4, the block DCT transform obtained to step 3;
Step 5 carries out mean value computation and slope fit processing to the local message entropy obtained in step 2 and step 4, is corresponded to
Average and slope;
Step 6 carries out the original image set of collection 2 times of down-sampling operations, and the image set after down-sampling is turned from rgb space
Lab space is changed to, operation of the step 2 to step 5 is repeated afterwards, obtains corresponding average and slope;
Step 7 carries out the original image set of collection 4 times of down-sampling operations, and the image set after down-sampling is turned from rgb space
Lab space is changed to, operation of the step 2 to step 5 is repeated afterwards, obtains corresponding average and slope;
Step 8, by the average extracted in step 5, step 6 and step 7 and slope characteristics and its corresponding air quality grade mark
Label are input to SVM classifier simultaneously and model to obtain Detection of Air Quality model;
Step 9, the model obtained with step 8 are detected air quality grade.
2. the air quality grade detection method according to claim 1 based on image, which is characterized in that will in step 1
The original image set of collection is transformed into Lab space from rgb space, is specially:
Step 1-1, image is transformed into XYZ space from rgb space, conversion formula is:
<mrow>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mi>X</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>Y</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>Z</mi>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>=</mo>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mn>0.4124</mn>
</mtd>
<mtd>
<mn>0.3575</mn>
</mtd>
<mtd>
<mn>0.1804</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0.2128</mn>
</mtd>
<mtd>
<mn>0.7152</mn>
</mtd>
<mtd>
<mn>0.0722</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0.0193</mn>
</mtd>
<mtd>
<mn>0.1192</mn>
</mtd>
<mtd>
<mn>0.9502</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mi>R</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>G</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>B</mi>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
Wherein, R, G, B distinguish the R passages of representative image, G passages, channel B, X, Y, Z represent respectively the X passage of image, Y passages,
Z passages;
Step 1-2, image is transformed into Lab space from XYZ space, conversion formula is:
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<msup>
<mi>t</mi>
<mrow>
<mn>1</mn>
<mo>/</mo>
<mn>3</mn>
</mrow>
</msup>
</mtd>
<mtd>
<mrow>
<mi>t</mi>
<mo>></mo>
<msup>
<mrow>
<mo>(</mo>
<mfrac>
<mn>6</mn>
<mn>29</mn>
</mfrac>
<mo>)</mo>
</mrow>
<mn>3</mn>
</msup>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mfrac>
<mn>1</mn>
<mn>3</mn>
</mfrac>
<msup>
<mrow>
<mo>(</mo>
<mfrac>
<mn>29</mn>
<mn>6</mn>
</mfrac>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mi>t</mi>
<mo>+</mo>
<mfrac>
<mn>4</mn>
<mn>29</mn>
</mfrac>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>t</mi>
<mo>&le;</mo>
<msup>
<mrow>
<mo>(</mo>
<mfrac>
<mn>6</mn>
<mn>29</mn>
</mfrac>
<mo>)</mo>
</mrow>
<mn>3</mn>
</msup>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
L=116f (Y/100) -16
A=500 [f (X/95.047)-f (Y/100)]
B=200 [f (Y/100)-f (Z/108.883)]
Wherein, t represents X, Y, Z passage, and f (t) is the transfer function from XYZ space to Lab space, and L, a, b represent image respectively
L passages, a passages, b passages.
3. the air quality grade detection method according to claim 1 based on image, which is characterized in that carried in step 2
The spatial domain local message entropy for taking image L passages is specially:
Step 2-1:Piecemeal processing is done to image L passages, local block size is 8*8;
Step 2-2:The comentropy of L path partially blocks is calculated, the calculation formula of comentropy is:
E=- ∑sxp(x)log2p(x)
Wherein, x refers to the value of pixel in L path partially blocks, and p (x) refers to the probability that pixel value is x in L path partially blocks,
E refers to localized mass information entropy;
Step 2-3:All spatial domain local message entropys of L passages according to comentropy size are ranked up, are expressed as:
<mrow>
<msub>
<mi>S</mi>
<mrow>
<mi>s</mi>
<mi>l</mi>
</mrow>
</msub>
<mo>=</mo>
<mrow>
<mo>(</mo>
<msubsup>
<mi>E</mi>
<mn>1</mn>
<mrow>
<mi>s</mi>
<mi>l</mi>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>E</mi>
<mn>2</mn>
<mrow>
<mi>s</mi>
<mi>l</mi>
</mrow>
</msubsup>
<mo>,</mo>
<mn>...</mn>
<mo>,</mo>
<msubsup>
<mi>E</mi>
<mi>n</mi>
<mrow>
<mi>s</mi>
<mi>l</mi>
</mrow>
</msubsup>
<mo>)</mo>
</mrow>
</mrow>
Wherein, n is spatial domain L path partially block numbers,The spatial-domain information entropy of n localized mass of L passages of finger
AndSslRepresent the spatial domain L path partially comentropy set after sequence.
4. the air quality grade detection method according to claim 1 based on image, which is characterized in that step 3 is to turning
The image for changing Lab space into does block DCT transform and is specially:
Localized mass dct transform is done respectively to image L, a, b passage, the calculation formula of dct transform is:
<mrow>
<mi>F</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>c</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>)</mo>
</mrow>
<mi>c</mi>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<mi>N</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msubsup>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<mi>N</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
<mi>c</mi>
<mi>o</mi>
<mi>s</mi>
<mo>&lsqb;</mo>
<mfrac>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mn>0.5</mn>
<mo>)</mo>
<mi>&pi;</mi>
</mrow>
<mi>N</mi>
</mfrac>
<mi>u</mi>
<mo>&rsqb;</mo>
<mi>c</mi>
<mi>o</mi>
<mi>s</mi>
<mo>&lsqb;</mo>
<mfrac>
<mrow>
<mo>(</mo>
<mi>j</mi>
<mo>+</mo>
<mn>0.5</mn>
<mo>)</mo>
<mi>&pi;</mi>
</mrow>
<mi>N</mi>
</mfrac>
<mi>v</mi>
<mo>&rsqb;</mo>
</mrow>
<mrow>
<mi>c</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<msqrt>
<mfrac>
<mn>1</mn>
<mi>N</mi>
</mfrac>
</msqrt>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>u</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msqrt>
<mfrac>
<mn>2</mn>
<mi>N</mi>
</mfrac>
</msqrt>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>u</mi>
<mo>&NotEqual;</mo>
<mn>0</mn>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
Wherein, f (i, j) is the pixel value that spatial domain L, a, b path partially block internal coordinate is (i, j), and local block size is N*N,
8*8 is taken, (u, v) is the coordinate of pixel after L, a, b path partially block dct transform, and F (u, v) refers to L, a, b path partially block
The DCT coefficient obtained after dct transform.
5. the air quality grade detection method according to claim 1 based on image, which is characterized in that step 4 carries out
Frequency domain local message entropy extracts:
Step 4-1, piecemeal processing is done to image L, a, b passage after block DCT transform, local block size is 8*8;
Step 4-2, the comentropy of image L, a, b path partially block after block DCT transform, the calculation formula of comentropy are calculated
For:
E=- ∑sxp(x)log2p(x)
Wherein, x refers to the value of pixel in image L, a, b path partially block after block DCT transform, and p (x) refers to single channel office
Pixel value is the probability of x in portion's block, and E refers to localized mass comentropy;
Step 4-3:It can be expressed as after sorting to all frequency domain local message entropys of L, a, b passage according to comentropy size:
<mrow>
<msub>
<mi>S</mi>
<mrow>
<mi>f</mi>
<mi>l</mi>
</mrow>
</msub>
<mo>=</mo>
<mrow>
<mo>(</mo>
<msubsup>
<mi>E</mi>
<mn>1</mn>
<mrow>
<mi>f</mi>
<mi>l</mi>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>E</mi>
<mn>2</mn>
<mrow>
<mi>f</mi>
<mi>l</mi>
</mrow>
</msubsup>
<mo>,</mo>
<mn>...</mn>
<mo>,</mo>
<msubsup>
<mi>E</mi>
<mi>n</mi>
<mrow>
<mi>f</mi>
<mi>l</mi>
</mrow>
</msubsup>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>S</mi>
<mrow>
<mi>f</mi>
<mi>a</mi>
</mrow>
</msub>
<mo>=</mo>
<mrow>
<mo>(</mo>
<msubsup>
<mi>E</mi>
<mn>1</mn>
<mrow>
<mi>f</mi>
<mi>a</mi>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>E</mi>
<mn>2</mn>
<mrow>
<mi>f</mi>
<mi>a</mi>
</mrow>
</msubsup>
<mo>,</mo>
<mn>...</mn>
<mo>,</mo>
<msubsup>
<mi>E</mi>
<mi>n</mi>
<mrow>
<mi>f</mi>
<mi>a</mi>
</mrow>
</msubsup>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>S</mi>
<mrow>
<mi>f</mi>
<mi>b</mi>
</mrow>
</msub>
<mo>=</mo>
<mrow>
<mo>(</mo>
<msubsup>
<mi>E</mi>
<mn>1</mn>
<mrow>
<mi>f</mi>
<mi>b</mi>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>E</mi>
<mn>2</mn>
<mrow>
<mi>f</mi>
<mi>b</mi>
</mrow>
</msubsup>
<mo>,</mo>
<mn>...</mn>
<mo>,</mo>
<msubsup>
<mi>E</mi>
<mi>n</mi>
<mrow>
<mi>f</mi>
<mi>b</mi>
</mrow>
</msubsup>
<mo>)</mo>
</mrow>
</mrow>
Wherein, n is frequency domain L, a, b path partially block number;The frequency domain information entropy of n localized mass of L passages of finger
AndSflRepresent the frequency domain L path partially comentropy set after sequence;Refer to a
The frequency domain information entropy of n localized mass of passage andSfaRepresent the frequency domain a path partially information after sequence
Entropy set;Refer to the frequency domain information entropy of n localized mass of b passages andSfbIt represents
Frequency domain b path partially comentropy set after sequence.
6. the air quality grade detection method according to claim 1 based on image, which is characterized in that step 5 is played a game
Portion's comentropy carries out mean value computation and slope fit processing, obtains corresponding average and slope, is specially:
Step 5-1, S is taken respectivelysl、Sfl、Sfa、Sfb60% local message entropy, value mode among four local message entropy set
For:
Step 5-2, respectively to SSl_60%、SFl_60%、SFa_60%、SFb_60%Four local message entropy set carry out mean value computations and tiltedly
Rate is fitted, and the average and slope calculation formula of local message entropy set are:
<mrow>
<msub>
<mi>mean</mi>
<mi>S</mi>
</msub>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mi>m</mi>
</mfrac>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>m</mi>
</msubsup>
<mi>S</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>skew</mi>
<mi>S</mi>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<mi>S</mi>
<mrow>
<mo>(</mo>
<mi>m</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mi>S</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
<mi>m</mi>
</mfrac>
</mrow>
Wherein, S represents SSl_60%、SFl_60%、SFa_60%、SFb_60%, m is the size of local message entropy set S, meanSIt is local letter
Cease the average of entropy set S, skewSRepresent the slope of local message entropy set S;
Step 5-3, the corresponding average of four local message entropy set and slope are expressed as:
F1=[meanSl_60%,skewSl_60%,meanFl_60%,skewFl_60%,meanFa_60%,skewFa_60%,meanFb_60%,
skewFb_60%]。
7. the air quality grade detection method according to claim 1 based on image, which is characterized in that step 6 obtains
Corresponding average and slope are:
<mrow>
<mi>f</mi>
<mn>2</mn>
<mo>=</mo>
<mo>&lsqb;</mo>
<msubsup>
<mi>mean</mi>
<mrow>
<mi>s</mi>
<mi>l</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>2</mn>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>skew</mi>
<mrow>
<mi>s</mi>
<mi>l</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>2</mn>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>mean</mi>
<mrow>
<mi>f</mi>
<mi>l</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>2</mn>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>skew</mi>
<mrow>
<mi>f</mi>
<mi>l</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>2</mn>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>mean</mi>
<mrow>
<mi>f</mi>
<mi>a</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>2</mn>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>skew</mi>
<mrow>
<mi>f</mi>
<mi>a</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>2</mn>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>mean</mi>
<mrow>
<mi>f</mi>
<mi>b</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>2</mn>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>skew</mi>
<mrow>
<mi>f</mi>
<mi>b</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>2</mn>
</mrow>
</msubsup>
<mo>&rsqb;</mo>
<mo>.</mo>
</mrow>
8. the air quality grade detection method according to claim 1 based on image, which is characterized in that step 7 obtains
Corresponding average and slope are:
<mrow>
<mi>f</mi>
<mn>3</mn>
<mo>=</mo>
<mo>&lsqb;</mo>
<msubsup>
<mi>mean</mi>
<mrow>
<mi>s</mi>
<mi>l</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>4</mn>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>skew</mi>
<mrow>
<mi>s</mi>
<mi>l</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>4</mn>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>mean</mi>
<mrow>
<mi>f</mi>
<mi>l</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>4</mn>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>skew</mi>
<mrow>
<mi>f</mi>
<mi>l</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>4</mn>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>mean</mi>
<mrow>
<mi>f</mi>
<mi>a</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>4</mn>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>skew</mi>
<mrow>
<mi>f</mi>
<mi>a</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>4</mn>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>mean</mi>
<mrow>
<mi>f</mi>
<mi>b</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>4</mn>
</mrow>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>skew</mi>
<mrow>
<mi>f</mi>
<mi>b</mi>
<mo>_</mo>
<mn>60</mn>
<mi>%</mi>
</mrow>
<mrow>
<mi>d</mi>
<mi>s</mi>
<mo>_</mo>
<mn>4</mn>
</mrow>
</msubsup>
<mo>&rsqb;</mo>
<mo>.</mo>
</mrow>
9. the air quality grade detection method according to claim 1 based on image, which is characterized in that SVM in step 8
Grader is realized by LibSVM kits.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711389016.6A CN108052980B (en) | 2017-12-21 | 2017-12-21 | Image-based air quality grade detection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711389016.6A CN108052980B (en) | 2017-12-21 | 2017-12-21 | Image-based air quality grade detection method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108052980A true CN108052980A (en) | 2018-05-18 |
CN108052980B CN108052980B (en) | 2021-05-14 |
Family
ID=62130725
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711389016.6A Active CN108052980B (en) | 2017-12-21 | 2017-12-21 | Image-based air quality grade detection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108052980B (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108985353A (en) * | 2018-06-27 | 2018-12-11 | 天津大学 | Based on the air pollution degree classification method that image is end-to-end |
CN109242821A (en) * | 2018-07-17 | 2019-01-18 | 深圳大学 | Air Quality Evaluation method, system, equipment and storage medium based on image quality evaluation |
WO2020014863A1 (en) * | 2018-07-17 | 2020-01-23 | 深圳大学 | Air quality assessment method and system based on image quality assessment, and device and storage medium |
CN110849879A (en) * | 2019-11-29 | 2020-02-28 | 华润电力技术研究院有限公司 | Carbon content detection method, carbon content detection system and computer storage medium |
CN111079848A (en) * | 2019-12-23 | 2020-04-28 | 哈尔滨理工大学 | Image-based air quality grade evaluation method |
CN111126493A (en) * | 2019-12-25 | 2020-05-08 | 东软睿驰汽车技术(沈阳)有限公司 | Deep learning model training method and device, electronic equipment and storage medium |
CN111177229A (en) * | 2019-12-11 | 2020-05-19 | 浙江工业大学 | Image-based indoor environment 3D dynamic monitoring method |
CN111504912A (en) * | 2020-05-11 | 2020-08-07 | 兰州大学 | Air pollution observation system and method based on image recognition |
CN112297768A (en) * | 2020-10-31 | 2021-02-02 | 中环曼普科技(南京)有限公司 | Vehicle air conditioner control method and system based on visual identification |
TWI777458B (en) * | 2021-03-18 | 2022-09-11 | 國立中正大學 | Air pollution detection method based on spectrum image |
US11499953B2 (en) | 2019-12-09 | 2022-11-15 | International Business Machines Corporation | Feature tuning—application dependent feature type selection for improved classification accuracy |
US11619618B2 (en) | 2019-12-09 | 2023-04-04 | International Business Machines Corporation | Sensor tuning—sensor specific selection for IoT—electronic nose application using gradient boosting decision trees |
CN117078778A (en) * | 2023-10-16 | 2023-11-17 | 四川邕合科技有限公司 | Intelligent park air quality detection method and detection terminal based on big data |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103048427A (en) * | 2012-12-25 | 2013-04-17 | 安科智慧城市技术(中国)有限公司 | Indoor air quality assessment method, device and system |
CN103218622A (en) * | 2013-04-22 | 2013-07-24 | 武汉大学 | Haze monitoring method based on computer vision |
CN103475898A (en) * | 2013-09-16 | 2013-12-25 | 北京理工大学 | Non-reference image quality assessment method based on information entropy characters |
WO2014174415A1 (en) * | 2013-04-26 | 2014-10-30 | Elica S.P.A. | Ventilation system for a room |
CN104517126A (en) * | 2014-12-26 | 2015-04-15 | 北京邮电大学 | Air quality assessment method based on image analysis |
US20150318015A1 (en) * | 2010-08-26 | 2015-11-05 | Blast Motion Inc. | Multi-sensor event detection system |
CN105493109A (en) * | 2013-06-05 | 2016-04-13 | 微软技术许可有限责任公司 | Air quality inference using multiple data sources |
CN106485281A (en) * | 2016-10-19 | 2017-03-08 | 葛嘉豪 | A kind of detection method of the Air haze class of pollution |
-
2017
- 2017-12-21 CN CN201711389016.6A patent/CN108052980B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150318015A1 (en) * | 2010-08-26 | 2015-11-05 | Blast Motion Inc. | Multi-sensor event detection system |
CN103048427A (en) * | 2012-12-25 | 2013-04-17 | 安科智慧城市技术(中国)有限公司 | Indoor air quality assessment method, device and system |
CN103218622A (en) * | 2013-04-22 | 2013-07-24 | 武汉大学 | Haze monitoring method based on computer vision |
WO2014174415A1 (en) * | 2013-04-26 | 2014-10-30 | Elica S.P.A. | Ventilation system for a room |
CN105493109A (en) * | 2013-06-05 | 2016-04-13 | 微软技术许可有限责任公司 | Air quality inference using multiple data sources |
CN103475898A (en) * | 2013-09-16 | 2013-12-25 | 北京理工大学 | Non-reference image quality assessment method based on information entropy characters |
CN104517126A (en) * | 2014-12-26 | 2015-04-15 | 北京邮电大学 | Air quality assessment method based on image analysis |
CN106485281A (en) * | 2016-10-19 | 2017-03-08 | 葛嘉豪 | A kind of detection method of the Air haze class of pollution |
Non-Patent Citations (2)
Title |
---|
BENQIAN YANG.ET: "PM2.5 Concentration Estimation Based on Image Quality Assessment", 《2017 4TH IAPR ASIAN CONFERENCE ON PATTERN RECOGNITION (ACPR)》 * |
许茜等: "融合空域与频域特征的能见度监督学习", 《计算机应用》 * |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108985353A (en) * | 2018-06-27 | 2018-12-11 | 天津大学 | Based on the air pollution degree classification method that image is end-to-end |
CN109242821A (en) * | 2018-07-17 | 2019-01-18 | 深圳大学 | Air Quality Evaluation method, system, equipment and storage medium based on image quality evaluation |
WO2020014863A1 (en) * | 2018-07-17 | 2020-01-23 | 深圳大学 | Air quality assessment method and system based on image quality assessment, and device and storage medium |
CN110849879A (en) * | 2019-11-29 | 2020-02-28 | 华润电力技术研究院有限公司 | Carbon content detection method, carbon content detection system and computer storage medium |
US11499953B2 (en) | 2019-12-09 | 2022-11-15 | International Business Machines Corporation | Feature tuning—application dependent feature type selection for improved classification accuracy |
US11619618B2 (en) | 2019-12-09 | 2023-04-04 | International Business Machines Corporation | Sensor tuning—sensor specific selection for IoT—electronic nose application using gradient boosting decision trees |
CN111177229A (en) * | 2019-12-11 | 2020-05-19 | 浙江工业大学 | Image-based indoor environment 3D dynamic monitoring method |
CN111079848A (en) * | 2019-12-23 | 2020-04-28 | 哈尔滨理工大学 | Image-based air quality grade evaluation method |
CN111079848B (en) * | 2019-12-23 | 2021-02-12 | 哈尔滨理工大学 | Image-based air quality grade evaluation method |
CN111126493A (en) * | 2019-12-25 | 2020-05-08 | 东软睿驰汽车技术(沈阳)有限公司 | Deep learning model training method and device, electronic equipment and storage medium |
CN111126493B (en) * | 2019-12-25 | 2023-08-01 | 东软睿驰汽车技术(沈阳)有限公司 | Training method and device for deep learning model, electronic equipment and storage medium |
CN111504912A (en) * | 2020-05-11 | 2020-08-07 | 兰州大学 | Air pollution observation system and method based on image recognition |
CN112297768A (en) * | 2020-10-31 | 2021-02-02 | 中环曼普科技(南京)有限公司 | Vehicle air conditioner control method and system based on visual identification |
TWI777458B (en) * | 2021-03-18 | 2022-09-11 | 國立中正大學 | Air pollution detection method based on spectrum image |
US11962949B2 (en) | 2021-03-18 | 2024-04-16 | National Chung Cheng University | Method of air pollution estimation based on spectral image processing |
CN117078778A (en) * | 2023-10-16 | 2023-11-17 | 四川邕合科技有限公司 | Intelligent park air quality detection method and detection terminal based on big data |
CN117078778B (en) * | 2023-10-16 | 2024-01-26 | 四川邕合科技有限公司 | Intelligent park air quality detection method and detection terminal based on big data |
Also Published As
Publication number | Publication date |
---|---|
CN108052980B (en) | 2021-05-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108052980A (en) | Air quality grade detection method based on image | |
CN103996192B (en) | Non-reference image quality evaluation method based on high-quality natural image statistical magnitude model | |
CN102799635B (en) | The image collection sort method that a kind of user drives | |
CN104134204B (en) | Image definition evaluation method and image definition evaluation device based on sparse representation | |
WO2016165082A1 (en) | Image stego-detection method based on deep learning | |
CN104408707B (en) | Rapid digital imaging fuzzy identification and restored image quality assessment method | |
CN101441717B (en) | Method and system for detecting eroticism video | |
CN104598885B (en) | The detection of word label and localization method in street view image | |
CN105046277A (en) | Robust mechanism research method of characteristic significance in image quality evaluation | |
CN104599280B (en) | A kind of crack on road crosspoint extracting method and system | |
CN107644415A (en) | A kind of text image method for evaluating quality and equipment | |
CN107808358A (en) | Image watermark automatic testing method | |
CN104778702A (en) | Image stego-detection method on basis of deep learning | |
CN106600595A (en) | Human body characteristic dimension automatic measuring method based on artificial intelligence algorithm | |
CN103475897B (en) | Adaptive image quality evaluation method based on distortion type judgment | |
CN111325687B (en) | Smooth filtering evidence obtaining method based on end-to-end deep network | |
CN103310235B (en) | A kind of steganalysis method based on parameter identification and estimation | |
CN102799872B (en) | Image processing method based on face image characteristics | |
CN106127741A (en) | Non-reference picture quality appraisement method based on improvement natural scene statistical model | |
CN106780727B (en) | Vehicle head detection model reconstruction method and device | |
CN106991666A (en) | A kind of disease geo-radar image recognition methods suitable for many size pictorial informations | |
CN104268590A (en) | Blind image quality evaluation method based on complementarity combination characteristics and multiphase regression | |
CN105374047B (en) | SAR image change detection based on improved bilateral filtering with cluster | |
CN106780446A (en) | It is a kind of to mix distorted image quality evaluating method without reference | |
CN107818321A (en) | A kind of watermark date recognition method for vehicle annual test |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |